Download - Cintolesi Makuc Thesis

Transcript
  • ABSTRACT

    Kinetic and Stoichiometric Modeling of the Metabolism of Escherichia coli for the

    Synthesis of Biofuels and Chemicals.

    by

    Angela Cintolesi

    This thesis presents the mathematical modeling of two new Escherichia coli

    platforms with economical potential for the production of biofuels and chemicals, namely

    glycerol fermentation and the reversal of the -oxidation cycle. With the increase in

    traditional fuel prices, alternative renewable energy sources are needed, and the efficient

    production of biofuels becomes imperative. So far studies have focused on using glucose

    as feedstock for the production of ethanol and other fuels, but a recent increase in

    glycerol availability and its consequent decrease in price make it an attractive feedstock.

    Furthermore, the reversed -oxidation cycle is a highly efficient mechanism for the

    synthesis of long-chain products. These two platforms have been reported experimentally

    in E. coli but their mathematical modeling is presented for the first time here.

    Because mathematical models have proved to be useful in the optimization of

    microbial metabolism, two complementary models were used in this study: kinetic and

    stoichiometric. Kinetic models can identify the control structure within a specific

    pathway, but they require highly detailed information, making them applicable to small

    sets of reactions. In contrast, stoichiometric models require only mass balance

    information, making them suitable for genome-scale modeling to study the effect of

    adding or removing reactions for the optimization of the synthesis of desired products.

  • To study glycerol fermentation, a kinetic model was implemented, allowing

    prediction of the limiting enzymes of this process: glycerol dehydrogenase and di-

    hydroxyacetone kinase. This prediction was experimentally validated by increasing their

    enzymatic activities, resulting in a two-fold increase in the rate of ethanol production.

    Additionally, a stoichiometric genome-scale model (GEM) was modified to represent the

    fermentative metabolism of glycerol, identifying key metabolic pathways for glycerol

    fermentation (including a new glycerol dissimilation pathway). The GEM was used to

    identify genetic modifications that would increase the synthesis of desired products, such

    as succinate and butanol.

    Finally, glucose metabolism using the reversal -oxidation cycle was modeled

    using a GEM to simulate the synthesis of a variety of medium and long chain products

    (including advanced biofuels). The model was used to design strategies that can lead to

    increase the productivity of target products.

  • Acknowledgement

    This thesis represents the result of four and a half years of work that would not

    have been possible if it was not for wonderful people that God has put on my path.

    First, I would like to thank my mom for her love, encouragement and support. I

    hope I will always live up to her vision: "hijos, el limite son las estrellas" ("children, the

    stars are the limit").

    I would like to thank my brothers and sisters, for their love and example.

    I would like to thank my advisor Dr. Ramon Gonzalez, for careful and patient

    advice and feedback throughout the years. This thesis would not have been possible

    without him.

    I would like to thank my thesis committee members, Dr. Luay Nakhleh and Dr.

    Deepak Nagrath, for their encouragement, feedback and support.

    I would like to thank my former professors from Rice and from Chile, who have

    believed in me and have encouraged me to be my very best.

    I would like to thank an enormous number of friends, who throughout the years

    have believed in me and supported my work.

    Last but not least, I want to express my gratitude to my Heavenly Father, for

    being the inspiration in my life, the one who has encouraged me the most to go to higher

    skies. The greatest scientist and the source of all good.

  • v

    Table of Contents

    1 Introduction ........................................................................................................ 1

    1.1 Motivation and justification ................................................................................. 1

    1.1.1 The problem of fuels and chemical production ............................................ 2

    1.1.2 Biofuels ......................................................................................................... 5

    1.1.3 Glycerol and glucose as feedstocks for ethanol and chemicals production .. 6

    1.2 Selection of microorganism: E. coli as platform .................................................. 8

    1.3 The use of mathematical modeling in the study of biological systems .............. 10

    1.4 Objectives ........................................................................................................... 11

    1.5 Organization of the thesis: Overview of the chapters ........................................ 12

    2 Background and literature review .................................................................. 14

    2.1 Metabolism of E. coli ......................................................................................... 14

    2.1.1 E. coli .......................................................................................................... 15

    2.1.2 Glycerol fermentation ................................................................................. 23

    2.1.3 Glucose utilization and the production of advanced biofuels ..................... 30

    2.2 Mathematical modeling of microbial metabolism ............................................. 37

    2.2.1 Stoichiometric modeling ............................................................................. 38

    2.2.2 Kinetic modeling ......................................................................................... 46

    2.2.3 Approximative kinetic models .................................................................... 51

    2.2.4 Remarkable models for E. coli central carbon metabolism and glycerol

    fermentation ............................................................................................................... 56

    3 Materials and methods..................................................................................... 64

    3.1 Modeling ............................................................................................................ 64

    3.1.1 Kinetic modeling and MCA ........................................................................ 65

    3.1.2 Stoichiometric modeling ............................................................................. 69

    3.2 Experimental work ............................................................................................. 72

    3.3 Workflow ........................................................................................................... 75

    4 Results I: Kinetic modeling and metabolic control analysis of the fermentative metabolism of glycerol fermentation. ..................................................... 76

    4.1 Model development and simulations.................................................................. 77

  • vi

    4.2 Parameters estimation: In vitro, in silico and optimization ................................ 86

    4.3 Results of modeling ............................................................................................ 89

    4.4 Metabolic control analysis of the fermentative utilization of glycerol .............. 92

    4.5 Experimental verification ................................................................................... 94

    4.6 Conclusions ........................................................................................................ 97

    5 Results II: Genome-scale modeling of the fermentative metabolism of

    glycerol and assessment of its potential as a platform for fuel and chemical

    production ........................................................................................................................ 99

    5.1 Model Implementation ..................................................................................... 101

    5.1.1 Define a starting GEM. ............................................................................. 101

    5.1.2 Definition of External Conditions. ............................................................ 102

    5.1.3 Model curation .......................................................................................... 103

    5.2 Results A: Understanding glycerol fermentation ............................................. 112

    5.2.1 Glycerol dissimilation ............................................................................... 114

    5.2.2 Essential Products ..................................................................................... 116

    5.2.3 Role of FHL .............................................................................................. 119

    5.2.4 Production of 3-C intermediate metabolites ............................................. 120

    5.2.5 Study of essential genes/reactions ............................................................ 122

    5.2.6 Targets Identified for Experimental Validation ........................................ 124

    5.2.7 Other Models ............................................................................................ 125

    5.3 Results B: Assessing the capabilities of glycerol fermentation as a platform for

    the synthesis of fuels and chemicals ........................................................................... 128

    5.3.1 1,2-PDO .................................................................................................... 131

    5.3.2 1,3-PDO .................................................................................................... 134

    5.3.3 Succinate ................................................................................................... 136

    5.3.4 D-Lactic acid ............................................................................................. 142

    5.3.5 Butanol ...................................................................................................... 144

    5.3.6 Propionic acid ........................................................................................... 146

    5.3.7 Propanol .................................................................................................... 149

    5.4 Conclusions ...................................................................................................... 152

    6 Results III: Genome scale model for the reversal of the -oxidation cycle 156

    6.1 Model implementation ..................................................................................... 158

  • vii

    6.1.1 Define a starting GEM .............................................................................. 158

    6.1.2 Define External Conditions....................................................................... 158

    6.1.3 Model curation .......................................................................................... 159

    6.1.4 Implementation of the reversal -oxidation pathway ............................... 161

    6.1.5 Implementation of FA biosynthesis pathway and -keto acid pathway ... 162

    6.1.6 Implementation of termination pathways ................................................. 163

    6.1.7 Implementation of pathways for the synthesis of products with a functional

    side chain ................................................................................................................. 166

    6.2 Results: Production of alcohols, alkanes and fatty acids ................................. 173

    6.2.1 Production of alcohol ................................................................................ 174

    6.2.2 Production of alkanes ................................................................................ 182

    6.2.3 Production of fatty acids ........................................................................... 187

    6.2.4 Comparison of engineered -oxidation reversal using ferredoxin- and

    NADH-dependent acyl-CoA dehydrogenases/trans-enoyl-CoA reductases. .......... 192

    6.2.5 Synthesis of products with functionalized side chain ............................... 194

    6.3 Conclusions ...................................................................................................... 203

    7 Final remarks and future directions............................................................. 206

    7.1 Summary of achievements presented in this thesis .......................................... 207

    7.2 Future directions ............................................................................................... 210

    8 'omenclature .................................................................................................. 214

    9 References ....................................................................................................... 219

    10 Appendix ......................................................................................................... 232

    10.1 Matrix system for FCC ..................................................................................... 232

    10.2 List of reactions modified in glyc-GEM .......................................................... 233

    10.3 List of reactions modified in lcp-GEM ............................................................ 237

    10.4 Single gene deletion ......................................................................................... 243

    10.5 Diagrams of new models identified by the glyc-GEM with high predicted

    specific growth rates.................................................................................................... 243

    10.6 Detail of calculations from experimental data for glycerol metabolism toward

    the synthesis of desired products. ................................................................................ 252

    10.7 Maximal theoretical yields for long chain products ......................................... 254

    10.8 Calculation of titers .......................................................................................... 262

  • viii

    10.9 Matlab function for the kinetic model .............................................................. 263

    10.10 Matlab function for glyc-GEM ..................................................................... 268

    10.11 Matlab function to test multiple models for glycerol fermentation .............. 273

    10.12 Matlab function for lcp-GEM ....................................................................... 279

    10.13 Matlab function for the synthesis of products derived from intermediate

    metabolites of the reversal -oxidation cycle .............................................................. 295

    10.14 Matlab function for the use of hydroxylated primers for the synthesis of

    carboxyacids and diols using reversal -oxidation cycle ............................................ 305

  • 1

    1 Introduction

    1.1 Motivation and justification

    The current petrochemical industry controls most of the production of

    transportation energy and other petrochemicals; however, this platform has several issues

    that make it necessary to look for alternative sources. The main issues are related to the

    instability of petroleum prices, the uncertainty of its availability in the future, and its

    negative impact in the environment. Although there are renewable sources to produce

    biofuels and biochemicals, which are environmentally friendlier, the price of using these

    alternative sources is not yet competitive with the traditional petrochemical industry.

    Therefore, there is an imperative necessity to continue exploring renewable sources to

    produce fuels and chemicals, as well as to optimize these processes.

    In this thesis, two new platforms for the production of renewable biofuels and

    chemicals in E. coli are studied using mathematical models: glycerol fermentation and

    the reversal of the -oxidation cycle. Glycerol can be consumed for E. coli in order to

    produce biofuels and biochemicals, and its use is promising when compared to other

    traditional carbon and energy sources for E. coli (glucose, for instance). The reversal of

    the -oxidation cycle is highly efficient in the synthesis of long chain products, such as

    advanced biofuels. A full study to increase the understanding of glycerol fermentation

    and the use of the reversal -oxidation cycle in E. coli, as well as to optimize and expand

    these mechanisms, requires mathematical modeling of the metabolism, a task that has not

    been fully explored yet and which is the focus of this thesis.

  • 2

    1.1.1 The problem of fuels and chemical production

    Our current society relies on the petrochemical industry for the production of

    transportation energy (fuels) and chemicals. Petrochemical products are those chemicals

    produced from petroleum, also called crude oil and natural gas. Crude oil and natural gas

    are a mixture of hydrocarbons, among them alkanes, cycloalkanes, and aromatic

    hydrocarbons. Petroleum is extracted and processed to obtain two main kinds of

    products: fuels and chemicals. Crude oil is extracted from the earth, and it is processed to

    separate the different components that originate valuable products. Petroleum products

    can be classified into two main groups: fuels and other derivatives (chemicals). Among

    the fuels are ethane, diesel fuel, fuel oils, gasoline, jet fuel, and kerosene. Some

    petroleum derivatives are alkenes, lubricants, wax, sulfuric acid, tar, asphalt, petroleum

    coke, paraffin wax, and aromatic petrochemicals. While petroleum derived fuels

    represent about the 94% of the source for total transportation energy (Reijnders &

    Huijbregts, 2009), derivative chemicals are used in the production of plastic, cosmetic,

    agrochemicals, adhesives, etc. But all this reliance on the petrochemical industry has

    achieved (and it will continue to achieve) unstable and even dangerous levels for society

    and the environment.

    The first concern is related to economic issues. Petroleum prices have been

    subject to enormous variation in past and present years. Few countries in the world

    produce petroleum, and these countries often have unstable political systems (Soetaert &

    Vandamme, 2009). The crisis in the 1970s, triggered by political issues in the Middle

    East, caused an enormous increase in the price of oil (Figure 1). In more recent years, the

    price of petroleum has continued to suffer huge increases and fluctuations. Before

  • 3

    September 2003, the price of crude oil was about $25/barrel (real price), and since then it

    has increased to over $100/barrel in 2011 (BPs Statistical Review of Worlds Energy

    Full Report 2012. Website: http://www.bp.com/statisticalreview. Figure 1 shows

    historical prices. Projections for the future are discouraging as, regardless of political

    issues, crude oil is a limited resource expected to run out in the near future.

    Figure 1: Crude oil prices 1861-2011. Statistical Review of World Energy June 2012. Website:

    http://www.bp.com/statisticalreview.

    Because petroleum is a non-renewable resource, and because of increasing

    demand, it is estimated that petroleum will run out in 50 years or less (Soetaert &

    Vandamme, 2009). Petroleum was formed by chemical processes that took place during

    millions of years on the earth, and therefore, it is a limited, non-renewable resource.

    Refineries were at first operating in areas of easy access. As easily accessed reserves are

    being exhausted, the process of extracting petroleum becomes more challenging. On the

    other hand, the forecasted demand of crude oil is increasing as a natural consequence of

    increases in population and in development. The U.S. Energy Information Administration

    120

    110

    100

    90

    80

    70

    60

    50

    40

    30

    20

    10

    1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010

    $ 2011

    $ money of the day

  • 4

    (Website: http://www.eia.gov) forecasts a steady projection in the global demand of

    liquid fuel, and an increase in the demand of liquid biofuels (Figure 2). In conclusion, the

    limited capacity of crude oil supply, together with the increase in crude oil price, creates

    an urgent necessity to develop alternative transport fuels.

    Figure 2: Forecast of global demand by fuel type (U. S. Energy Information Administration. Independent

    Statistics and Analysis. AEO2013 Early Release Overview. Website: http://www.eia.gov)

    But petroleum does not have just economical and availability issues, it is also

    unsafe to the environment. The petrochemical industry has been related to CO2

    emissions, air pollution, and acid rain (Union of Concerned Scientists Clean Energy.

    Website: http://www.ucsusa.org). Carbon dioxide is produced by the combustion of fuels

    and is related to global warming. Although there are multiple sources for carbon dioxide

    emissions, studies reveal that these emissions have increased about 35% when compared

    to pre-industrial times (Environmental Protection Agency. Website: http://www.epa.gov),

    suggesting a relationship with the increased use of fuels. According to the U.S.

    department of energy, CO2 emissions related to energy are connected to over 80% of the

    greenhouse emissions (See Figure 3 for detail of other sources). Air pollution is another

  • 5

    negative side effect of using petroleum, causing damage to the health of humans. Burning

    fossil fuel generates several harmful emissions, among them carbon monoxide, nitrogen

    oxides and sulfur oxides. These emissions are related to health issues such as headaches,

    irritation of lungs, bronchitis, and respiratory infections. Finally, some of these emissions

    cause additional damage to the environment, such as acidic rain. Acidic rain has a

    negative impact on water, aquatic animal life, soils, and vegetation.

    Figure 3: Greenhouse gas emissions by gas, 2009 (Emission of greenhouse gases in the United States 2009. U.S

    Energy Information Administration, Dec 2009. Website: ftp://ftp.eia.doe.gov)

    1.1.2 Biofuels

    The expected increase in the price of crude oil, together with the inevitable

    running out of it and the environmental damage associated with this industry, has lead

    several countries to make efforts in incorporating alternatives. Although different kinds

    of renewable energy are being explored, among them nuclear energy, solar energy, wind

    power, and biofuels, the last one is the most appropriate for transportation (Zidansek et

    al., 2009). For instance, solar and wind power are not very reliable as they depend on

    external conditions, and nuclear energy still has some safety issues that would delay any

  • 6

    implementation for several years (Zidansek et al., 2009). In contrast, biofuels have

    proven to be appropriate for transport energy and some countries are already

    incorporating them. In addition, biofuels are made from biomass, which can also provide

    substitutes for some derivatives of the petrochemical industry, making biofuels even

    more attractive as a replacement of fossil products.

    Production of biofuels, and specifically of ethanol and biodiesel, was triggered by

    economic and environmental factors. After the oil crisis of 1973, Brazil decided to

    incorporate ethanol as an energy source for transport, and now it is one of the countries

    with the highest production of ethanol in the world (Leland, 2009; Reijnders &

    Huijbregts, 2009). Following the example of Brazil, the U.S. implemented a program to

    stimulate the use of ethanol in 1978, and is currently the other main producer of ethanol

    in the world (Leland, 2009; Reijnders & Huijbregts, 2009). The U.S. Department of

    Energy expects to quadruple the consumption of renewable biofuels in the coming years,

    from 9 billion gallons in 2008 to 36 billion ethanol-equivalent gallons in 2022 (U.S.

    Department of Energy, 2012), and other countries (such as Canada, South Africa and

    Germany) are also implementing similar strategies to replace part of traditional fuel

    consumptions (Reijnders & Huijbregts, 2009).

    1.1.3 Glycerol and glucose as feedstocks for ethanol and chemicals production

    Traditionally feedstocks for the production of biofuels have been cane and corn,

    for the production of ethanol, and oil crops, for production of biodiesel (Fischer et al.,

    2008). Cane and corn contain sucrose and starch respectively, which can be broken down

    into sugars, mainly glucose and fructose, and then be fermented to produce ethanol.

    Sugar fermentation is a relatively simple process that can take place in yeast and other

  • 7

    organisms. However, in order to make biofuels economically viable, it is necessary to

    develop new and more efficient processes to use glucose and other less expensive

    biofuels. One way to achieve this is in the use of alternative feedstock, such as cellulose,

    syngas, fatty acids, and glycerol, all of which have been proposed as alternatives to

    reduce cost in the production of biofuels (Dellomonaco et al., 2010; Fischer et al., 2008;

    Yazdani & Gonzalez, 2007). Another way to make biofuels more attractive economically

    is to concentrate on the production of advanced biofuels, which contain higher energy

    density than traditional fuels. In this thesis two approaches are studied: 1) the use of

    glycerol as a cheaper carbon source to produce biofuels and chemicals and 2) the use of a

    newly discovered metabolic pathway, the reversal of the -oxidation cycle, to produce

    advanced biofuels and other chemicals. The reason for this selection is explained in the

    following two paragraphs.

    Glycerol is a 3-carbon molecule that has two important and promising properties

    for the production of biofuels: low price and a high degree of reduction (Yazdani &

    Gonzalez, 2007). Glycerol is a by-product in the production of biodiesel, and its

    production is 10% (weight) of the biodiesel product. The increase in production of

    biodiesel in recent years has lead to a 10- fold decrease in the price of glycerol between

    2004 and 2006 (Yazdani & Gonzalez, 2007), and its current price is even lower ($0.005

    per pound. 01/2013; www.thejacobsen.com). The low price of glycerol makes it

    competitive when compared to sugars. Furthermore, the highly reduced nature of carbon

    atoms in glycerol (higher than sugars) makes it even more attractive for the production of

    biofuels. For instance, glycerol presents a maximum theoretical yield twice as large as the

    one produced when glucose is used as the carbon source in the anaerobic fermentation in

  • 8

    E. coli (Dharmadi et al., 2006). Regardless of experimental efforts that have been

    conducted to improve the utilization by E. coli of glycerol for the productions of

    biofuels and chemicals, no mathematical modeling has been conducted previous to

    this thesis, a task that is imperative for further understanding and optimization of

    this process.

    The reversal of the -oxidation cycle is a newly engineered process that results in

    the efficient synthesis of long chain chemicals, including advanced biofuels. While the -

    oxidation cycle is the most commonly used metabolic pathway for the utilization of fatty

    acids as energy and carbon sources, the reversal of this process is possible by

    metabolically engineering the pathway. It experimental implementation in E. coli using

    glucose as carbon and energy source was demonstrated in 2011 by Dellomonaco and

    collaborators (Dellomonaco et al., 2011). The implementation of this pathway resulted in

    the production of long chain linear alcohols and fatty acids at higher yields than those

    reported using other pathways, a highly the attractive aspect of this pathway

    (Dellomonaco et al., 2011). In order to increase the understanding of the reversal of

    the -oxidation cycle in E. coli as well as to explore its capabilities to enhance the

    optimal production of a variety of long chain chemicals, the implementation of

    mathematical model is essential, a task that has been undertaken in this thesis.

    1.2 Selection of microorganism: E. coli as platform

    The production of biofuels requires a platform capable of transforming the

    feedstock into the desired product. The complexity of this operation, together with the

    high efficiency that biological systems provide when compared to chemical catalysts,

    resulted in the decision to use biological microorganisms, which have proven to be able

  • 9

    to synthesize biochemicals. In addition, several tools have been developed in the field of

    biotechnology during the last 40 years (starting with the discovery of restriction enzymes

    in the 1970s), focused on the modification of wild strains in order to improve their

    original properties. Biotechnological tools allow, for instance, modifying the tolerance of

    cells to external conditions, adding or removing reactions from the original pathway in

    order to increase the production of the desired product, overexpressing or

    underexpressing native genes, and modifing proteins characteristics (reversibility,

    activity, among others).

    Although different microorganisms are able to produce biofuels, this study

    focuses on one in particular: Escherichia coli. E. coli is a gram negative bacterium that

    has been widely studied for applications in biotechnology. Several properties make E.

    coli an ideal organism for the production of biochemicals on an industrial scale. First, E.

    coli can easily grow in attractive industrial conditions. It can grow with and without

    oxygen, use several substrates, and use inexpensive media components (Clomburg &

    Gonzalez, 2010). Second, the large amount of information on the metabolism of these

    bacteria allows efficient manipulation of it using gene transformation, regulation of gene

    expression, and protein engineering (Stephanopoulos, 2007). Furthermore, the

    development of omics technology (genomics, transcriptomics, and proteomics) has

    been applied extensively to E. coli, allowing an increase in the understanding of this

    bacterium. This accumulated knowledge has allowed mathematical modeling of part of

    the metabolism of E. coli, the remarkable examples being the central carbon metabolism

    for glucose utilization (Chassagnole et al., 2002) and genome-scale modeling (Orth et al.,

    2011).

  • 10

    Most importantly, recent studies confirm that E. coli is able to ferment glycerol in

    the absence of external electron acceptors, an ability shared only by a few organisms

    (Dharmadi et al., 2006), which can lead to the production of ethanol other products. In

    addition, E. coli can produce long chain chemicals (including advanced biofuels) utilizing

    a newly engineered pathway: the reversal of the -oxidation cycle (Dellomonaco et al.,

    2011).

    1.3 The use of mathematical modeling in the study of biological systems

    Prior to the utilization of mathematical models in biological systems, biological

    parts (genes, proteins, reactions, and metabolites) were studied in a limited way with

    laboratory technology. Most of the experiments consisted of the modification of one

    single element (a gene, for instance) and then the observation of the effect over other

    components of the system. This approach has changed in recent years, and several efforts

    have been made to apply mathematical models to the study of biological systems as a

    whole.

    The use of mathematical models allows a deeper understanding of the roles of

    different parts of the system, and how their manipulation can increase the production of a

    desired product under specific conditions. For example, a mathematical model can allow

    the identification of the limiting steps in a system, while otherwise it would be necessary

    to performance multiple experiments in the laboratory, which are expensive and time

    consuming, to reach the same conclusion. Gombert and Nielsen reviewed the two main

    approaches in modeling metabolic pathways: stoichiometric models and kinetic models

    (Gombert & Nielsen, 2000). Furthermore, some models incorporate the genetic

  • 11

    regulatory network in order to represent different conditions that suppose the expression

    of different genes (see for example Ramakrishna et al., 1996 and Moisset et al., 2012).

    1.4 Objectives

    The main goal of this thesis is the development of mathematical models for the

    understanding of microbial utilization of glycerol and the implementation of a

    functional reversal of the -oxidation in E. coli related to the production of biofuels

    and other valuable chemicals. This was done using two approaches: kinetic modeling

    and genome-scale modeling. In particular, the following tasks were performed:

    Implementing a kinetic model for the fermentative metabolism of glycerol.

    This model was used to elucidate the control structure in this pathway, to create

    predictions, and to propose genetic modifications for increasing the production of

    biofuels.

    Implementing a genome-scale (stoichiometric) model (GEM) for the

    fermentative metabolism of glycerol. This model allowed the identification of

    active pathways and engineered pathways for the efficient conversion of glycerol

    into biofuels and other chemicals.

    Implementing a GEM for the utilization of glucose to produce long chain

    chemicals using the reversal of the -oxidation cycle. This model allowed for

    the evaluation of different strategies for the production of a variety of advanced

    biofuels and other long chain chemicals. These strategies include the evaluation of

    adding and removing reactions.

  • 12

    1.5 Organization of the thesis: Overview of the chapters

    The next section of this thesis is Background and literature review. That

    section starts by reviewing metabolic information available for E. coli, which includes

    detailed studies of pathways related to utilization of glycerol and glucose. Specific

    pathways for the consumption of glycerol and glucose are explained. Different pathways

    that have been proposed to produce advance biofuels and other chemicals are explained,

    including the -oxidation cycle as well as its engineered reversal for the production of

    long chain products. Then, mathematical models for metabolic pathways are presented,

    including approximate models and applications.

    Two mathematical methods were selected for this study: kinetic modeling and

    GEM. Materials and Methods explains why these methods were selected and gives

    further details of their implementations. Also, detailed experimental protocols are

    presented for the validation of predictions and other relevant findings.

    The next three sections present Results, which consist of 1) a kinetic model for

    glycerol fermentation, 2) a GEM for glycerol fermentation, and 3) a GEM for glucose

    metabolism for the production of medium and long chain products using the reversal of

    the -oxidation cycle. The kinetic model accounts for the glycerol transport and

    dissimilation, glycolysis, and ethanol synthesis, and it represents correctly the

    experimental data. This model was used to study the control structure of glycerol

    fermentation and predictions were experimentally validated. The GEM for glycerol in E.

    coli accurately represents glycerol fermentation. Implementation of this model allowed

    increasing the understanding of glycerol fermentation, as well as investigating the

    synthesis of additional products using optimal strategies. The GEM for the reversal of the

  • 13

    -oxidation cycle was used to evaluate this pathway as an efficient platform for the

    production of a variety of medium and long chain products, including advanced biofuels,

    and it increases the understanding of redox balance and energy requirements of this

    pathway. In addition, a GEM was used to simulate the use of different platforms for the

    production of medium and long chain products. It was concluded that the reversal of the

    -oxidation cycle predicts higher productivities based on a higher energetic efficiency.

    To conclude, Final Remarks and Future Directions presents a summary of the

    main findings exposed in this thesis, as well as suggestions about future experimental

    work to corroborate the findings of this work. In future experimental work it is suggested

    that a variety of genetic modifications be implemented, leading to i) corroboration of the

    existence of new unstudied pathways and ii) production of biofuels and chemicals at

    optimal levels. Some future directions in the use of mathematical models in metabolic

    engineering, in particular en the area of production of biofuels and chemicals are also

    suggested.

  • 14

    2 Background and literature review

    This chapter describes fundamentals for building mathematical models for

    glycerol fermentation and respiratory utilization of fatty acids in E. coli. First, the

    metabolism of E. coli is described, with special emphasis on glycerol and glucose

    utilization. The fermentative metabolism of glycerol and glucose includes glycolysis and

    production of fermentative products, among others. The problem of achieving redox

    balance is explained, and how this affects both the cell growth and the maximum

    theoretical yield for biofuels production. Then, the use of mathematical tools to model

    metabolism is explained, including different approaches that have been used. A special

    emphasis is given to models that have been successfully applied to E. coli and to models

    for glycerol fermentation that have been developed in other organisms. Then, the

    advantages and challenges of producing advanced biofuels and other long chain

    chemicals are presented, including the newly reported reversal of the -oxidation cycle.

    This provide a comprehensive vision of the different efforts to model the metabolism of

    E. coli and the utilization of glycerol and glucose for the synthesis of biofuels and

    chemicals.

    2.1 Metabolism of E. coli

    The cellular metabolism consists of all reactions that uptake nutrients from the

    environment, and then utilize them for cell function and reproduction. Metabolism can be

    divided into two main processes: catabolism and anabolism. The catabolism includes all

    those reactions that break down large molecules in order to obtain energy and building

    blocks. Building blocks are used in the anabolic pathways in order to produce

  • 15

    macromolecules that enable the maintenance and reproduction of cells, such as proteins,

    polynucleotides (DNA and RNA) and lipids. Energy and redox power are two important

    concepts that govern the metabolism, as they are involved in a vast number of pathways.

    2.1.1 E. coli

    This study utilizes the bacterium E. coli K-12, on whose metabolism a vast

    number of studies have been performed. These studies cover both specific pathways and

    genome scale studies. While studies of the metabolism started from studies of particular

    pathways, such as glycolysis and TCA cycle, genome scale studies attempt to include all

    reactions in the metabolism. This section first explains available genome scale

    information for this microorganism and then it explains specific pathways related to the

    utilization of glycerol and fatty acids.

    Genome scale studies

    Prior to the publication of the genome scale project for E. coli, just 1,853 genes

    had been identified in this organism. In 1997 the genome scale project revealed that E.

    coli has 4,288 ORFs (potential genes) (Blattner et al., 1997), and this number has been

    updated to 4,499 genes (Keseler et al., 2013). ORFs potentially encode information for

    proteins, whose main functional categories are metabolism (enzymes), transportation

    (across cellular membranes), regulation, and structure. The database EcoCyc collects the

    information related to each of these ORFs of E. coli and presents it in a user-friendly and

    integrated way (Keseler et al., 2013). Both sequence analysis and experiments are used to

    assign the biochemical function of a gene product (Karp et al., 2007). By 2006 EcoCyc

    curators finished performing a literature review of each of the ORFs of E. coli, and they

    keep updating the database (Karp et al., 2007; Keseler et al., 2013). Among the genes that

  • 16

    E. coli has, 76% of them have a biochemical function assigned, and in most of those

    cases there are experiments that support the function (Karp et al., 2007). In addition,

    EcoCyc present all known metabolic pathways of E. coli, and it also presents related

    information for each enzyme in a reaction, such as activators, inhibitors, and cofactors.

    By 2012, EcoCyc cited 23,909 distinct references, and included 300 metabolic pathways,

    which is considerably more than the 194 metabolic pathways reported by this site in 2007

    (Karp et al., 2007). A classification of these pathways is complicated given the

    complexity of the metabolic network, but a general classification considers biosynthetic

    pathways, degradation pathways, detoxification, generation of precursor metabolites and

    energy, and signal transduction pathways (taken from EcoCyc website).

    The information presented in EcoCyc has been used and compared with the

    metabolic model built by the group of Palsson (Karp et al., 2007). The final result of this

    effort is a GEM, named the iJO1366, that accounts for all known metabolic pathways

    presented in EcoCyc (Orth et al., 2011). In addition, this model includes spontaneous

    reactions that are not catalyzed by enzymes, and therefore they are not part of the EcoCyc

    database. The model includes 2,251 reactions and 1,366 ORFs, representing the effort of

    21 years working on the reconstruction of the complete metabolic network of E. coli

    (Figure 1).

  • 17

    Figure 4: Seven milestones efforts in the construction of metabolic network of E. coli (Feist & Palsson, 2008).

    This figure was updated to include the information from the last released GEM for E. coli, the iJO1366 (Orth et

    al., 2011).

    Another relevant source of information in the study of the metabolism of E. coli is

    the BRaunschweig ENzyme DAtabase, most commonly referred to as BRE'DA

    (Schomburg et al., 2013). This database collects detailed enzymatic information, such as

    biochemical and molecular information, from nearly 500 organisms. As with EcoCyc,

    BRENDA curators collect information from literature. BRENDA is not specific to any

    organism, but for each enzyme, it offers the option to restrict the displayed information to

    a specific organism, among them E. coli. This database is especially useful for collecting

    kinetic information, as it links to articles that have reported rate law, reversibility, kinetic

    parameters, and possible inhibition of the corresponding reaction.

    Genomic Era

    1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

    Majewski

    & Domach

    Varma &

    Palsson

    Pramanik

    & Keasling

    Edwards

    & Palsson

    Reed

    et al.

    Feist et

    al.

    Orth et

    al.

    0

    250

    500

    750

    1,000

    1,250

    2,250

    2,500

    Reactions

    Genes

    Metabolites

    Amino acids

    and nucleot.

    biosyn.

    Cell wall constituent biosyn.

    Growth-depen. biomass obj. func.

    Cofactor biosynthesis

    Fatty acids metabolism

    Expanded cellular transport systems

    Used genome as a scaffold

    Alternative carbon utilization

    Quinone characterization

    Elemental and charge balancing

    Compartmentalized reconstruction (ditinct periplasm)

    Extensive cell wall metabolism (phospholipids, murein, LPS)

    Reactions thermodynamics

    Updates in genes, metabolites and reactions

    Nu

    mb

    er

  • 18

    Glycolysis and fermentative pathways

    Pathways in the central metabolism are especially relevant in this study, as they

    allow the utilization of different carbon sources, such as glucose, malate, succinate, and

    acetate. Glucose and other sugars have been widely studied, as they represent an efficient

    cellular growth. Anaerobic fermentation of glucose uses the glycolysis and fermentative

    pathways in order to obtain energy (in the form of ATP), and it results in the production

    of lactate, ethanol, and other fermentative products (Sawers & Clark, 2004). In the case

    of aerobic utilization of glucose, the tricarboxylic acid cycle (TCA cycle) is also

    incorporated, which results in a higher production of ATP and therefore a more efficient

    cell growth. Glycolysis and fermentative pathways are explained in the next paragraphs.

    Glycolysis is for the most part a linear pathway that starts from the metabolite

    glucose-6-phosphate (G6P) and culminates in the production of pyruvate (Figure 5).

    Glucose can be transported and phosphorylated into G6P by the phosphotransferase

    system (PTS), a well studied mechanism that is also related to catabolic repression of the

    utilization of other carbon sources (Romeo & Snoep, 2005). Other carbon sources, such

    as galactose and maltose, are first transformed in an intermediate metabolite, glucose-1-

    phosphate (G1P), which is then converted into G6P by the action of a

    phosphoglucomutase (Pgm) (Romeo & Snoep, 2005). The first reaction in glycolysis is

    the conversion of G6P into fructose-6-phosphate (F6P) mediated by phosphoglucose

    isomerase (Pgi). F6P is an essential metabolite that either goes to the pentose phosphate

    pathway (PP pathway) or continues in the glycolytic pathway by adding a second

    phosphorylation that results in fructose-bi-phosphate (FBP). The interconversion of F6P

    and FBP is considered a regulatory step in the glycolysis, as it is catalyzed by two

  • 19

    phosphofructokinases (Pfk-1 and Pfk-2) that present allosteric and genetic regulation

    (Romeo & Snoep, 2005). Pfk-1 and Pfk-2, are encoded by the genes pfkA and pfkB

    respectively, the first being responsible for 90% of the enzymatic activity in this step

    (Romeo & Snoep, 2005). FBP is reversibly dissociated into two metabolites,

    glyceraldehyde-3-phosphate (GAP) and di-hydroxyacetone-phosphate (DHAP). DHAP

    can be transformed into GAP by the action of a triosephosphate isomerase (TPI) and

    continue with the glycolytic pathway, or it can be transformed into the toxic metabolite

    methylglyoxal (MG). GAP is interconverted through four successive reactions into

    phosphoenolpyruvate (PEP), resulting in the production of one molecule of ATP and one

    molecule of NADH (equivalent to two reducing equivalent, H) (Romeo & Snoep, 2005).

    The last reaction in glycolysis is the conversion of PEP into pyruvate (PYR), producing

    one ATP. This reaction is catalyzed by the action of two pyruvate kinases (PykF and

    PykA), both subject to genetic and allosteric regulation (Romeo & Snoep, 2005). In the

    case of glucose utilization, glycolysis represents the production of two NADH (four

    reducing equivalents, H) given that one molecule of glucose allows the production of two

    molecules of PYR. These molecules of NADH need to be reoxidized to maintain redox

    poise, which is achieved either by using fermentative pathways or by the action of an

    external electron acceptor.

    In the absence of external electron acceptors, E. coli utilizes fermentative

    pathways to generate products that are more reduced than the starting metabolite,

    resulting in a pathway that oxidizes the reducing equivalents generated during the

    glycolysis. Fermentative pathways can produce a variety of products, such as acetate,

    carbon dioxide, dihydrogen, ethanol, formate, lactate, and succinate (Sawers & Clark,

  • 20

    2004). Figure 5 shows how these fermentative pathways connect to the glycolysis,

    starting from either PEP or PYR.

    Figure 5: Glycolysis and fermentative pathways for E. coli. Intermediate metabolites are shown in black and

    fermentative products appear in green. (Sawers & Clark, 2004).

    In order to understand how the fermentative pathways allow the cell to achieve

    redox balance, it is necessary to consider the degree of reduction of carbon molecules in

    substrates, products, and biomass production. The degree of reduction per carbon in a

    molecule can be calculated in three steps: 1) define a set of reference compounds with a

    G

    L

    Y

    C

    O

    L

    Y

    S

    I

    S

    F

    E

    R

    M

    E

    N

    T

    A

    T

    I

    O

    N

  • 21

    degree of oxidation equal to zero, 2) calculate the oxidation number of each element in

    the reference system, and 3) calculate the degree of reduction by adding the number of

    oxidation in the molecule and dividing it by the amount of carbons (Ratledge &

    Kristiansen, 2001). Selection of H2O, H+, HCO3

    -, SO42-, and H2SO4 as reference

    compounds results in the following oxidation number per element: O = -2, H = 1, C = 4,

    S = 6, and N = -3. Table 1 shows the degree of reduction per carbon in substrates,

    metabolites, products, and biomass. Thus, the 2 molecules of NADH generated in the

    glycolysis during glucose utilization can be reoxidized by transforming PYR into some of

    the products with a higher degree of reduction, such as ethanol or lactate. The amount of

    reducing equivalent that each of the fermentative pathways generates is shown in Figure

    5.

    Table 1: Oxidation state and degree of reduction of various substrates, intermediate, products, biomass and

    reactions.

    Compound Formula Degree of reduction

    Substrates

    Glucose C6H12O6 4

    Xylose C5H10O5 4

    Glycerol C3H8O3 4.67

    Intermediate

    PYR C3H3O3 3

    Products

    Acetic Acid C2H4O2 4

    Ethanol C2H6O 6

    Formic acid CH2O2 2

    Lactic acid C3H6O3 4

    Succinic acid C4H6O4 3.5

    1,2-PDO C3H8O2 5.33

    1,3-PDO C3H8O2 5.33

    Carbon dioxide CO2 0

    Biomass CH1.9O0.5N0.2 4.3

  • 22

    Although different fermentative pathways allow achieving redox balance, not all

    of them are efficient in energy production. For instance, one molecule of PYR going to

    ethanol consumes 2 molecules of NADH and does not produce energy; in contrast,

    production of acetate from PYR produces one molecule of ATP, but it does not consume

    any reducing equivalent. Since glucose utilization allows the production of two molecules

    of PYR, producing 50% ethanol and 50% acetate is an option that both achieves redox

    balance and produces ATP.

    So far the analysis of redox balance has been done without including the

    production of biomass. The degree of reduction of reactions can be calculated as the

    difference of degree of reduction of products and substrates, weighted by the

    stoichiometric number and the amount of carbons. Thus, if glucose is used for the

    production of biomass, it results in the consumption of part of the reducing equivalents

    generated during glycolysis. On the other hand, if the carbon source is more reduced than

    the biomass (e.g. glycerol), the fermentative pathways need to be able to reoxidize the

    reducing equivalents generated during glycolysis and biomass production.

    In the presence of an external electron acceptor, such as oxygen, fermentative

    pathways are no longer required to achieve redox balance, and metabolites generated in

    the glycolysis can go to the TCA cycle. This cycle is part of the aerobic respiration

    process, and it generates energy (as ATP) by reducing metabolites and subsequently

    increasing the number of reducing equivalents (Alberts, 2002). The final electron

    acceptor in respiration is oxygen, which allows the reoxidation of reducing equivalents.

  • 23

    2.1.2 Glycerol fermentation

    Glycerol fermentation is an attractive system to produce biofuels because of its

    potentially high efficiency. When compared with common sugars such as glucose and

    xylose, the highly reduced nature of glycerol (Table 1) represents a higher theoretical

    yield of ethanol (Dharmadi et al., 2006). This section presents the current knowledge of

    glycerol fermentation in bacteria.

    Glycerol fermentation in bacteria

    Among the few species that are able to ferment glycerol anaerobically, there are

    Klebsiella, Enterobacter, and Citrobacter. The mechanism that allows these bacteria to

    ferment glycerol has been studied for years. In these organisms, the fermentation of

    glycerol is linked to the ability to metabolize glycerol through two pathways: reductive

    and oxidative (Lin, 1976; Zhu et al., 2002). Figure 6 shows the two pathways that allow

    the utilization of glycerol: the oxidative pathway produces DHAP and the reductive

    pathway produces 1,3-propanediol (1,3-PDO). In the oxidative pathway, glycerol is

    dehydrogenated by a NAD-dependent glycerol dehydrogenase (glyDH-I), resulting in the

    formation of dihydroxyacetone (DHA). Then, DHA is phosphorylated by a DHA kinase

    (DHAK), resulting in the formation of DHAP, which is incorporated to the glycolysis.

    DHAK requires either PEP or ATP to donate the phosphate group (Yazdani & Gonzalez,

    2007). Additionally, fermentation of glycerol in these members of the

    Enterobacteriaceae family is linked to their ability to produce 1,3-PDO. Production of

    1,3-PDO starts with the dehydration of glycerol by the coenzyme B12-dependent glycerol

    dehydrogenase (GlyD) resulting in the production of 3-hydroxypropionaldehyde (3-

  • 24

    HPA). This molecule is then reduced to 1,3-PDO by the action of the NADH-dependent

    1,3-PDO dehydrogenase (1,3-PDODH) (Yazdani & Gonzalez, 2007).

    Figure 6: Glycerol utilization model for species of the Enterobacteriaceae. 1,3-PDO: 1,3-propanediol, GLYC:

    glycerol, DHA: dihydroxyacetone, 3HPA: hydroxypropionaldehyde, DHAP: di-hydroxyacetone phosphate,

    PYR: pyruvate, DHAK: di-hydroxyacetone kinase (Yazdani & Gonzalez, 2007).

    To understand why glycerol fermentation requires the production of 1,3-PDO in

    members of the Enterobacteriaceae family, it is necessary to understand how to achieve

    redox balance. First, the degree of reduction of glycerol and biomass (shown in Table 1)

    indicates that glycerol is more reduced than biomass (4.67 versus 4.3), which is the

    opposite case than in the utilization of glucose. Consequently, biomass production

    starting from glycerol generates reducing equivalent. In addition, production of ethanol is

    redox balanced when glycerol is the carbon source (Figure 5). Although the ethanol

    fermentation pathway is able to consume the reducing equivalents generated during

    uptake of glycerol and glycolysis, this pathway is not able to consume any of the extra

    reducing equivalents that would be generated during cell growth; therefore, a different

    pathway is necessary. Production of 1,3-PDO allows the reoxidation of reducing

    equivalents but this pathway does not generate energy; consequently, both pathways

    (production of ethanol and 1,3-PDO) need to be combined to enable maintenance and cell

    growth.

  • 25

    A few other organisms have been reported to ferment glycerol without producing

    1,3-PDO, such as Propionibacteria freudenreichii and Propionibacteria acidipropionici

    ssp shermanii, (Bories et al., 2004). Nevertheless, the pathways that enable these

    organisms to ferment glycerol have not been well studied.

    Glycerol fermentation in E. coli

    E. coli does not have the ability to produce 1,3-PDO; consequently, it was

    believed that this organism fermented glycerol only in the presence of an external

    electron acceptor. Aerobic utilization of glycerol starts with the transport of glycerol into

    the cell using passive diffusion or facilitated diffusion by the aquaglyceroporin GlpF

    (Heller et al., 1980). The gene encoding this enzyme (glpF) belongs to the same operon

    as the gene glpK. The product of glpK is an ATP-dependent glycerol kinase (GlpK) that

    catalyzes the conversion of glycerol into glycerol-3-phosphate (G3P) (Booth, 2005). Two

    glycerol 3-phosphate dehydrogenases (GlpD and GlpABC, for aerobic and anaerobic

    conditions respectively) allow the oxidation of G3P into DHAP (Booth, 2005). These

    enzymes are membrane-bound flavin-dependent, and the reaction results in the

    production of one reduced ubiquinone (ubiquinol). In the absence of external electron

    acceptors, the second reaction cannot proceed as there are no later steps to reoxidize the

    ubiquinol, resulting in the accumulation of G3P in toxic levels (Booth, 2005). Figure 7

    shows this pathway.

  • 26

    Figure 7: Respiratory utilization of glycerol by E. coli. PYR: pyruvate, G3P: glycerol-3-phosphate, DHAP: di-

    hydroxyacetone phosphate. G3P: glycerol-3-phosphate, an-G3PDH: anaerobic G3P dehydrogenase, ae-G3PDH:

    aerobic G3P dehydrogenase (Murarka et al., 2008).

    The discovery that E. coli is able to consume glycerol in the absence of external

    electron acceptors motivated several studies to understand the experimental conditions as

    well as the pathways required in this process, since E. coli is a highly attractive host for

    the production of chemicals and biofuels. The first experiment that showed fermentative

    utilization of glycerol in E. coli utilized a rich medium (Dharmadi et al., 2006), but later

    experiments were performed in a minimum medium supplemented with tryptone, sodium

    selenite, and Na2HPO4 (Murarka et al., 2008). In these batch experiments, the initial

    concentration of glycerol was 110 mM at 37C. The optimum pH was identified as 6.3,

    although alkaline conditions were tolerable when a supplemented minimum medium was

    used. Fermentation of near 80% of the glycerol took 96 hours, with a maximum specific

    growth rate of 0.04 h-1 (Murarka et al., 2008). Figure 8 shows the utilization of glycerol

    in 96 hours, and the corresponding production of biomass, ethanol and other byproducts.

    Products of this process are 1,2-propanediol (1,2-PDO), acetic acid, succinic acid, formic

    acids (minor amounts), carbon dioxide, hydrogen, and ethanol, the latter corresponding to

    the most abundant (other than hydrogen and carbon dioxide) (Gonzalez et al., 2008;

    Murarka et al., 2008). Ethanol corresponds to nearly 95% of the fermentative products,

    while 1,2-PDO presents a very low concentration (0.5 0.15 mM in stationary phase)

  • 27

    (Gonzalez et al., 2008). These results contrast with the products of glucose fermentation

    in E. coli, where the main product is lactate (not observed in glycerol fermentation),

    ethanol represents only a 29%, and there is no evidence of 1,2-PDO production (Sawers

    & Clark, 2004).

    Figure 8: Glycerol fermentation by E. coli MG1655. Cell density: black squares. Cell density log-linear: white

    squares. Ethanol: black circles. Glycerol: triangles. Succinic acid: rhombus. Formic plus acetic acids: cross. 1

    OD = 0.34 gCDW/L.

    Glycerol fermentation in E. coli starts with the conversion of glycerol into DHA

    catalyzed by the enzyme glycerol dehydrogenase (glyDH), a type two glycerol

    dehydrogenase encoded by the gene gldA (Gonzalez et al., 2008). The existence of this

    gene was known before the discovery of E. coli being able to ferment glycerol, but it was

    considered cryptic (Truniger & Boos, 1994). The discovery that E. coli can express this

    enzyme is key in the fermentation of glycerol, and it constitutes one of the essential

    enzymes of this process (Gonzalez et al., 2008). The next step is the phosphorylation of

    DHA into DHAP by the action of the PEP-dependent di-hydroxyacetone kinase (DHAK).

    This enzyme, likewise glyDH, is essential for the utilization of glycerol in the absence of

    external acceptors (Gonzalez et al., 2008).

  • 28

    Studies were conducted to investigate the effect of alternative pathways for the

    breakdown of glycerol, as well as the branches to different products. Disruption of the

    gene glpK (responsible for the conversion of glycerol into G3P) decreased the maximum

    specific growth rate in 17.5% (Murarka et al., 2008), suggesting that although this

    enzyme is not essential, it contributes to the process. In addition, deletion of fermentative

    pathways indicates that ethanol production is essential for glycerol fermentation, while

    disruption of acetate production (deletion of gene pta) resulted in a decrease in the

    maximum specific growth rate of nearly 30% (Murarka et al., 2008). Disruption of the

    enzyme formate-hydrogen lyase (FHL), responsible for the conversion of formate into

    hydrogen (H2) and carbon dioxide (CO2), results in a significant reduction of cell growth,

    a phenomenon than has been attributed to the necessary presence of CO2 in the media and

    the negative effects of H2, as it affects the redox balance (Gonzalez et al., 2008; Murarka

    et al., 2008).

    Given the inability of E. coli to produce 1,3-PDO, and the finding that 1,2-PDO is

    a byproduct of glycerol fermentation, a new model linked to 1,2-PDO production was

    proposed to explain the fermentation of glycerol (Gonzalez et al., 2008). Figure 10

    presents two pathways that could allow the production of 1,2-PDO in E. coli. They both

    start with the conversion of DHAP into MG, a highly toxic compound for E. coli (Booth,

    2005). MG can be converted into 1,2-PDO via production of either hydroxyacetone (HA)

    or lactaldehyde (Lald), both reducing steps that result in the consumption of one NADH

    (or NADPH). Experimental data support the production of 1,2-PDO through HA rather

    than through Lald, as genetic disruption in the first branch decreases the production of

    1,2-PDO to nearly 65% of original values, while disruption in the second branch does not

  • 29

    have any effect (Gonzalez et al., 2008). Interestingly, the enzyme that catalyzes the

    reduction of HA to 1,2-PDO is glyDH (encoded by gldA), the same that oxidizes glycerol

    to DHA in the first step of glycerol breakdown under fermentative conditions, which

    highlights the importance of this enzyme.

    Figure 9: Metabolic pathways for the production of 1,2-PDO in E. coli, starting form DHAP. Tick lines

    represent the most probably pathway for glycerol fermentation in E. coli (Gonzalez et al., 2008). DHAP: di-

    hydroxyacetone phosphate, MG: methylglyoxal, HA: hydroxyacetone, Lald: lactaldehyde, 1,2-PDO: 1,2

    propanediol.

    Figure 10 presents the glycerol uptake pathway, combining with the production of

    fermentative products and 1,2-PDO. The transformation of DHA to DHAP via a PEP-

    dependent (DHAK) requires that this reaction combines with the last step in the

    glycolysis, which is responsible for the conversion of PEP into PYR. A simple

    stoichiometric analysis predicts that PEP is converted into PYR at the same rate that

    DHA is converted to DHAP, but this is incompatible with the fact that part of the flux is

    going to the production of 1,2-PDO. One pathway that could partially explain it is the

    conversion of glycerol into DHAP going through G3P, a pathway that requires the

    already mentioned GlpK and GlpABC. Nevertheless, disruption of this pathway only

    results in a decrease in the cell growth, not in a total inability to ferment glycerol.

    Therefore, these experimental findings indicate that additional pathways may play an

    important role in the fermentation of glycerol in E. coli, a fact that emphasizes the

    Lald

  • 30

    importance of conducting additional studies in glycerol fermentation in order to identify

    all pathways involved in the process. Despite experimental studies of glycerol

    fermentation in E. coli, no mathematical analysis of glycerol fermentation has been

    conducted previous to this thesis.

    Figure 10: Proposed model for glycerol fermentation in E. coli. Production of ethanol and 1,2-PDO is essential to

    support cell growth as a combined synthesis of these two products supports the generation of ATP and the

    consumption of reducing equivalents. GldA: glycerol dehydrogenase, DHAK: di-hydroxyacetone kinase, FHL:

    formate hydrogen lyase, PFL: pyruvate formate lyase, AdhE: alcohols/aldehyde dehydrogenase (Gonzalez et al.,

    2008).

    2.1.3 Glucose utilization and the production of advanced biofuels

    The mechanism that carries the consumption of glucose in E. coli has been well

    documented, as glucose is a preferred substrate by this and many other microorganisms.

    The incorporation of glucose into E. coli utilizes the phosphotransferase system (PTS),

    which results in the transport and phosphorylation of glucose to produce G6P using PEP

    as the phophoryl group-donor (Mayer & Boos, 2005). The PTS consist of the general

  • 31

    component and the sugar-specific component. The general component includes the

    enzyme I (EI), encoded by the gene ptsI, and the histidine protein (HPr), encoded by the

    gene ptsH. The sugar-dependent component includes the enzyme complexes II,

    corresponding to enzymes EIIA, encoded by the gene crr, and EIIBC, encoded by the

    gene ptsG. The phosphoryl group is subsequently transferred from PEP to EI, HPr, EIIB,

    and then to glucose, resulting in the production of G6P. The domain EIIC is responsible

    for the transport of glucose across the membrane into the cell (Mayer & Boos, 2005).

    Once G6P is in the cytoplasmatic compartment, it can be metabolized in the glycolysis as

    explained above (page 18), or go into the pentose phosphate pathway (PPP) to produce

    the essential metabolites ribose-5-phosphate, sedpheptulose-7-phosphate and erythrose-4-

    phospthate (Ecocyc website).

    The PTS system is also involved with the gene expression regulation system. In

    the absence of glucose, the predominantly present phosphorylated EIIA causes catabolic

    repression of genes involved in the transport and utilization of sugars. In contrast, the

    presence of glucose increases the concentration of nonpnosphorylated EIIA, which

    prevents the use of other substrates by inhibiting the activity of non-PTS sugar transport

    system (Mayer & Boos, 2005).

    Glucose has been utilized for the production of advanced biofuels in E. coli.

    Advanced biofuels are molecules with a higher energy density than traditional biofuels,

    making them more comparable to traditional fuels. For example, ethanol, the most widely

    used biofuels, only contains 70% of the energy density of gasoline, while butanol, an

    advance biofuel, contains 84% of the energy density of gasoline (Atsumi et al., 2008a;

    Peralta-Yahya & Keasling, 2010). However, for most advanced biofuels there is no

  • 32

    natural pathways to produce them (butanol being the exception), but in recent years a

    number of efforts have been made in experimental work to implement the production of

    advanced biofuels in bacteria, and specifically in E. coli. The following paragraphs

    review the most remarkable pathway to convert glucose into advanced biofuels: the fatty

    acid (FA) biosynthesis pathway, the -keto acid pathway, and the reversal of the -

    oxidation cycle.

    E. coli has been metabolically engineered to produce long chain chemicals using

    the FA biosynthesis pathway and proper termination enzymes (Steen et al., 2010).

    The FA biosynthesis pathway is a well studied process in E. coli, in which one molecule

    of AcCoA is converted into malonyl-ACP (malACP) using energy, and then the malACP

    is utilized to elongate an acyl-ACP molecule (Cronan & Rock, 2008). A diagram of this

    pathway is shown in Figure 11. The mechanisms that converts AcCoA into malACP

    requires the genes accC, accAD, acpH, acpS and fabD, and it uses the energy of one

    phosphoryl group from ATP (Figure 11-A). The molecule malACP can be combined with

    one molecule of AcCoA (using enzyme FabH) to produce acetoacety-ACP, which initiate

    FA synthesis, or it can be used as carbon donor in the elongation cycle (Figure 11-B).

    The elongation cycle for FAs consists of four reactions: reduction (catalyzed by enzyme

    FabG), dehydrogenase (catalyzed by enzyme FabZ), reduction (catalyzed by enzyme

    FabI), and elongation (catalyzed by enzymes FabB and FabF) (Cronan & Rock, 2008).

  • 33

    Figure 11: FA biosynthesis pathway. Panel A shows the production of malACP and acetoacetyl-ACP for

    elongation and initiation of the production of fatty acids. Panel B shows the elongation cycle, which incorporates

    the two carbon molecules from mal-ACP to produce acyl-ACP (Cronan & Rock, 2008).

    Using the FA biosynthesis pathway and metabolic engineering, Steen and

    collaborators proved the production of a variety of long chain products, such as fatty

    alcohols and fatty esters (Steen et al., 2010). Figure 12 shows a diagram that summarizes

    this pathway and the genetic modifications involved. Synthesis of desired products

    resulted from the overexpresion of thioesterases (TES) and acyl-CoA ligases (ACL) to

    produce AcCoA, and the different products were obtained by overexpressing fatty-acyl-

    CoA reductase (FAR) for the production of fatty alcohols, and acyltransferase (AT) for

    the production of esters. Other products were also produced in this study, including

    biodiesels. Production of long chain FAs was obtained at a yield of 6% (w/w), while

    production of long chain fatty alcohols was not reported in yields, a 2% (w/v) of glucose

    was converted into 60 mg/l, which correspond to a yield of 3% (w/w) assuming that all

    the glucose was consumed (not reported) (Steen et al., 2010).

    A B

  • 34

    Figure 12: Metabolic pathway for the production of long chain products using the FA biosynthesis pathway.

    Figure is an adaptation from Steen and collaborators. TES: Thioesterase, ACL: Acyl-CoA ligases, FAR: fatty-

    acyl-CoA reductase, pdc: pyruvate decarboxylase, adhB: alcohol dehydrogenase, AT: Acyltransferase, Pyr:

    pyruvate, EtOH: ethanol, AcAld: acetaldehyde (Steen et al., 2010).

    A different approach that has been implemented for the production of synthesis of

    long chain products is the -keto acid pathway, which unlike the FA biosynthesis

    pathway, incorporates one molecule of carbon per cycle. This pathway was developed in

    Dr. James Liao's research center, and it modifies an aminoacids biosynthesis pathway for

    the production of linear alcohols. Figure 13 shows a diagram of how this pathway

    converts glucose into butanol. Using this strategy the production of butanol consists of

    using the biosynthesis pathway for the production of threonine, converting this molecule

    into 2-ketobutyrate, and then elongating 2-ketobutyrate incorporating one molecule of

    acyl-CoA and releasing one molecule of CO2 ( utilizing genes leuA, leuCD and leuB),

    resulting in the production of the intermediate 2-ketavalerate. Finally, production of

    butanol is achieved by reducing 2-ketovalerate using genes kivd (from Lactococcus

    lactis) and ADH2 (from Saccharomyces cerevisiae) (Figure 13) (Shen & Liao, 2008).

    Thus, the combination of gene leuA, leuCD and leuB allows the elongation in one

    FAs-derived

    productsGlucose

  • 35

    molecule of carbon, and it extension to produce hexanol has been proved (Marcheschi et

    al., 2012).

    Figure 13: Schematic illustration of the production of propanol and butanol utilizing the threonine biosynthesis

    pathway (Shen & Liao, 2008). Enzymes LeuA, LeuCD and LeuB elongate the intermediate product in one

    carbon, and it can be utilized to produce 1-pentanol and 1-hexanol.

    More recently a new promising pathway has been implemented for the synthesis

    of long chain products in E. coli: the reversal -oxidation cycle. In 2011, Dellomonaco

    and collaborators engineered the -oxidation cycle, which normally breaks down fatty

    acids into AcCoA molecules, to produce medium and long chain products (includying

    alcohols and fatty acids) starting from glucose (Dellomonaco et al., 2011). The

    implementation of this pathway required the modification of the regulatory pathway in

    wild type E. coli in order to activate this pathway in the absence of natural substrate (i.e.,

    fatty acids) and the presence of glucose. The metabolic pathway requires the breakdown

  • 36

    of glucose into AcCoA, a molecule that is used both as starting point for the reversal -

    oxidation cycle, and as carbon donor for the elongation step in this cycle (Figure 14).

    Figure 14: Illustration of the reversal -oxidation cycle to synthesize long chain products. Glucose is first break

    down into acetyl-CoA in the glycolysis, a process that results in production of ATP and reducing equivalents.

    Then acetyl-CoA is utilized to produce the primer of the cycle, and also as elongation molecule in the cycle.

    Expression of proper termination enzymes results in the production of alcohols and fatty acids (Dellomonaco et

    al., 2011).

    The primer of the reversal -oxidation cycle is synthesized by the enzyme acetyl-

    CoA acetyltransferase (AtoB), which catalyze the conversion of two molecules of acetyl-

    CoA into one molecule of acetoacetyl-CoA (a ketoacyl-CoA). The reversal -oxidation

    cycle has 4 steps in addition to the starting point: elongation, reduction (first),

    dehydration and reduction (second). Ketoacyl-CoA molecule is reduced by the action of

    FadB, producing a molecule of hydroacyl-CoA. The reaction oxidizes one molecule of

    NADH in turn. Hydroacyl-CoA is dehydrated by the action of FadB. A second reduction

    occurs by the action of enoyl-CoA reductase, a reaction that has been proposed to use

    ferredoxins as reducing equivalents, associated to the gene ydiO. The expression of

  • 37

    proper termination enzymes resulted in the production of long chain FAs at yields of 28%

    (w/w), and of higher chain liner alcohols (C6-C10) at 8.3% (w/w). These values are

    higher than those reported using other the FA biosynthesis pathway and the -keto acid

    pathway, which was proposed to be due energetic efficiencies of this pathway

    (Dellomonaco et al., 2011). More recently additional experimental efforts have been done

    using this pathway, especially in order to have a more controllable system for the

    regulatory system (Clomburg et al., 2012), but previous to this thesis no studies have

    been done using mathematical models. The utility of these models, as well as the state of

    art of pertinent mathematical models, is the focus of the following section of this chapter.

    2.2 Mathematical modeling of microbial metabolism

    Mathematical modeling of metabolism in microbial organisms consists in

    expressing in mathematical language the interaction between biochemical reactions,

    metabolites, substrates, and products. Biochemical reactions are those reactions that

    interconvert metabolites (or substrates) into other metabolites (or products). Because

    these reactions take place in a microorganism, they are often catalyzed by enzymes. From

    a structural point of view, there are two main categories of mathematical modeling of

    metabolism: stoichiometric models and kinetic models (Gombert & Nielsen, 2000).

    These two models differ in complexity and applicability, and they provide answers to

    questions such as what is the transient answer of the system when an external variable is

    changed, which are the control reactions in a specific pathway, what is the maximum

    theoretical yield of biomass, and what pathways need to be activated or deleted in order

  • 38

    to increase product yield. These two kinds of models are complementary to each other as

    they study the metabolism from different points of view.

    2.2.1 Stoichiometric modeling

    Stoichiometric models require the definition of a set of metabolites and the

    reactions that interconnect these metabolites. Each reaction should keep the law of

    conservation of mass; in other words, there must be a valid stoichiometric relationship in

    each reaction. In addition to this, stoichiometric models assume that the system is in a

    steady state, and therefore there is no accumulation of any internal metabolite. A

    stoichiometric matrix (S) contains all the stoichiometric parameters, in which rows

    represent reactions and columns represent metabolites. Because this matrix is used for

    mathematical calculus, linearly dependent cofactors (such as ATP and ADP) are included

    just once in the stoichiometric matrix in order to avoid having a linearly dependent

    matrix. Figure 15 represent the implementation of stoichiometric modeling for a metabolic

    network. In steady a state, it is valid to say that the net rate of formation of each

    metabolite minus the dilution rate associated with that metabolite ( ) is equal to zero (Figure 15, c). The dilution term is usually neglected because levels of intracellular

    metabolites are small; therefore, this flux is also small when compared with other fluxes

    of production and consumption of metabolites (Stephanopoulos et al., 1998). This leads

    to an equation system, in which the vector of fluxes v contains the unknown rates (Figure

    15, c). Nevertheless, the system is often underdetermined and additional tools should be

    included.

  • 39

    Figure 15: Stoichiometric Modeling. (a) Shows a segment of the structure of a metabolic network that is

    modeled. The metabolic network is translated into a matrix (b) in which rows represent metabolites and

    columns correspond to reactions. Finally, the stoichiometric matrix is used to write a system of equations in

    steady-state (c).

    Stoichiometric models are used to find a list of feasible reactions that satisfy

    steady state conditions, but it is necessary to include more information in order to come

    near to realistic solutions. Because of this, stoichiometric models are often combined with

    Metabolic Flux Analysis (MFA), which incorporates measurements of fluxes (external

    and/or internal) in order to reduce the degrees of freedom of the system. The application

    of MFA to a metabolic pathway has been explained by Stephanopoulos et al., and a

    summary of that method is presented here (Stephanopoulos et al., 1998).

    Depending on the system and on the number of measured fluxes, the system

    might be determined, underdetermined, or overdetermined. Suppose that there are K

    metabolites and J fluxes in the system. In the matrix form, this means that the dimension

    of v is Kx1 and the dimension of S is JxK. The second equation in Figure 15, c presents a

    system with a degree of freedom F = J K. To have a determined system, we need to

    measure F independent fluxes, which will reduce the freedom of the system to zero.

    Mi-2

    Mi

    Mi+1

    Mi-1

    ri-2

    ri

    ri-1

    (a) (b)

    (c)

    1 0 1 0 1 1 0 0 1

    Mi-2 Mi-1 Mi

    S =ri-2

    ri

    ri-1

    ST v = rmet = 0

    rmet - M= 0

  • 40

    Assume F fluxes were measured and storage in the vector vm, and the vector vc storages

    the remaining unknown fluxes. Then the second equation in Figure 15, c can be re-written

    as:

    0 = = + (1) which lead to a unique solution for the unknown fluxes:

    = (2) The matrix can be inverted since this is a square matrix of dimension KxK,

    and the solution is unique. If the system is overdetermined (the number of measured

    fluxes is greater than J K) and there is little noise in measurements, the system can be

    determined using the Moore-Penrose pseudo-inverse matrix of , named #, which is square matrix. In this case we have:

    = # , (3) where # can be calculated as:

    # = . (4) If the system is underdetermined, optimization might be combined with MFA in

    order to elucidate the fluxes. In stoichiometric modeling, the system is linear; therefore,

    all theory from linear optimization can be applied. Here is presented a brief description of

    linear optimization, but it will be further explained together with Flux Balance Analysis

    (FBA). A linear optimization problem requires the definition of an objective function and

    a set of constraints. The objective function is usually the optimization of cell growth

    (Stephanopoulos et al., 1998), although other functions have also been proposed, such as

    optimization of product formation. Then, the system is restricted to the space of valid

    solutions that are allowed for each flux rate.

  • 41

    MFA requires the incorporation of measured fluxes. There are two ways to do

    this: measurement of external fluxes (without labeling) and measurement of internal

    fluxes (requires labeling). When this information is incorporated, the degrees of freedom

    of the system decrease. Measurement of internal metabolites increases the complexity of

    experiments but it also increases the quality of the calculus. In order to do this, substrates

    are labeled and then the labeled state of internal metabolites is measured. One widely

    used method for the determination of intracellular fluxes is the incorporation of isotopes

    of carbon (13C or 14C) in a specific position of a substrate. This results in an introduction

    of an asymmetry in the distribution of labeled carbon, which can be detected and used to

    calculate internal fluxes (Stephanopoulos et al., 1998). Distribution of internal

    metabolites is measured using either gas chromatography-mass spectrometry (GC-MS) or

    Nuclear Magnetic Resonance (NMR).

    Stoichiometric modeling has also been used in conjunction with flux balance

    analysis (FBA). This method uses optimization in order to simulate microbial

    metabolism (Kauffman et al., 2003). The fundamental difference between MFA and FBA

    is the scope: while MFA focuses on the determination of metabolic fluxes for one

    experimental condition, the goal of FBA is to predict metabolic fluxes of a system under

    new conditions (Llaneras & Pico, 2008). Optimization of a linear problem requires, as

    was already stated, the definition of an objective function and a set of constraints for the

    unknown fluxes. The objective function is a linear combination of other fluxes in the

    system, and the objective must either maximize or minimize that function. The

    mathematical formulation of this is as follows:

    !"# $%&!"#'& = ()*(+ - ./ , (5)

  • 42

    where J is the number of fluxes of the system and i is the coefficient associated with the

    flux . The objective in FBA is usually the maximization of biomass production, and other possible objective functions are the maximization of ATP production and

    maximization of a desired product (Kauffman et al., 2003). The use of Biomass as

    Objective Function (BOF) has been validated experimentally; in fact, FBA applied to E.

    coli using BOF has shown a quantitative prediction of growth in 86% (68 out of 79) of

    mutants examined by Edwards and Palsson (Edwards & Palsson, 2000). Constraints refer

    to the range of valid values that each flux can take. For instance, those reactions that are

    known to be irreversible should be restricted to the range of positive values. In general

    this requirement can be written in the following mathematical form:

    0 % where 0 and % represent the lower and upper admissible values for flux . Other constraints that can be included are regulatory constraints (which fluxes are allowed

    under specific circumstances) (Covert et al., 2001) and incorporation of thermodynamics

    constraints (Beard et al., 2002).

    Once the metabolic network has been established in conjunction with the

    stoichiometric relationship, flux constraints, and an objective function, the problem can

    be solved using a mathematical software. Figure 16 represents a hypothetical case in

    which and 2 are unknown fluxes. Each of these fluxes has been restricted to a positive value, and the maximum value is as shown in the figure. Inclusion of additional

    restrictions coming from stoichiometric relationships defines the space of admissible

    solutions, also called solution space (area in color). The objective function is a linear

  • 43

    combination of fl